iiinova — where your digital self keeps evolving

Resting brains have uniformly high ID, and tasks selectively suppress ID — globally in infants, but with increasing precision through childhood and targeted compression by adulthood. This suggests that specialization isn't built by adding complexity where needed, but learning where to pare it away.

052

Tracking this across childhood and adolescence, we found this specificity emerges early and grows logarithmically with age — and it reflects task-irrelevant regions learning to compress their activity, rather than task-relevant regions expanding.

141

This coupling is completely absent in infants — they show reliable stimulus-driven responses but uniformly low, undifferentiated dimensionality everywhere — no evidence yet of "spending" complexity where it's needed.

120

In adults, brain regions that reliably track a story or movie also show higher-dimensional activity — a tight coupling between task engagement and representational complexity, across both auditory and visual stimuli.

110

Erica Busch @elbusch.bsky.social · 15/06/2026

🧠 New preprint! How does the brain build specialized, efficient representations as we grow up? We used manifold learning to track the "intrinsic dimensionality" (ID) of brain activity in ~800 participants (aged 3mo–53yrs), as they performed naturalistic tasks and rested/slept.

14315

Also excited to chat with anyone interested in learning more about my future lab at Vanderbilt launching in 2027! Please email me for more info

030

Erica Busch @elbusch.bsky.social · 15/06/2026

Excited to be at #ohbm2026 @ohbmofficial.bsky.social presenting our real-time fMRI paper just published in @natneuro.nature.com ! come learn more at my talk on Wed at 9am in the symposium on “Closed-loop fMRI Neurofeedback” in Room E

1143

A brain-computer interface that works with, not against, the brain

This project started in 2020 & shaped my PhD, so it's especially exciting to see it out. Massive thanks to my coauthors @glajoie.bsky.social & @krishnaswamylab.bsky.social & everyone who helped collect this massive dataset. news.yale.edu/2026/06/09/brain-computer-interface-works-not-against-brain

news.yale.edu

In a new study, Yale researchers developed a new kind of brain-computer interface that lets humans rapidly learn how to play a video game with their brains.

0111

For motor/communication disorders, manifold-aligned interfaces could work for far more people. For mental health, incremental manifold-aware approaches may outperform blunter interventions. More broadly: understanding the natural structure of our minds may be key to improving them.

230

Accordingly, the manifold determines both what people can learn and how fast. Our findings dovetail beautifully with work across species and recording methods, pointing to broad principles of how the brain learns. The manifold isn't just a neuroscience curiosity — it has real implications.

141

BCI learning changed the brain itself. Neural activity reorganized to align with the interface's goals & the degree of reorganization predicted individual differences in learning. Effects rippled beyond the targeted regions — suggesting BCI learning reshapes neural architecture more broadly.

141

When the mapping respected the manifold, people learned fast. When it didn't, learning stalled entirely — no improvement in the same amount of training time.

152

We designed three BCI mappings: one following the brain's most natural patterns, one using less dominant but still natural patterns, and one requiring patterns the brain doesn't naturally produce.

140

Multi-view manifold learning of human brain-state trajectories - Nature Computational Science

Brain activity doesn't wander randomly through high-D space. It travels along well-worn routes, forming a low-D structures or neural manifold. Using T-PHATE, which we developed previously (nature.com/articles/s43588-023-00419-0), we learned each person's individual manifold prior to BCI training.

nature.com

A manifold learning method called T-PHATE is developed for high-dimensional time-series data. T-PHATE is applied to brain data (functional magnetic resonance imaging) where it faithfully denoises sign...

191

BCIs have been around for decades, but about 1/3 of users never gain control at all, regardless of practice. We think this struggle isn't a facet of effort, but a design problem: BCIs asked the brain to generate activity that's entirely unnatural & don't provide the support for improvement.

130

Human learning of noninvasive brain–computer interfaces via manifold geometry - Nature Neuroscience

Our new paper is out this week in Nature Neuroscience! www.nature.com/articles/s41... We built a BCI that works with the brain's natural geometry — and we found that people could learn to play a video game with their brains in <1 hr of training. This efficiency is groundbreaking & here's why:

nature.com

Busch et al. use nonlinear neural manifolds to help humans gain rapid control over a noninvasive brain–computer interface, allowing them to learn how to play a video game with real-time fMRI neurofeed...

419777

I aim pay forward all of their support, optimism, and humor to my future mentees. Which brings me to: my lab at Vanderbilt CCC will be recruiting folks of all levels for Fall 2027! If you'll be at OHBM, come to my talk on 6/17 to get a sense of what we'll be up to — or just shoot me an email 🧠 🥳

162

I feel incredibly privileged to have a brilliant and generous team of mentors behind me - from my PhD advisor Nick Turk-Browne to my many mentors: @krishnaswamylab.bsky.social, Arielle Baskin-Sommers, BJ Casey, and @haxbylab.bsky.social. I cannot express enough gratitude to them & so many others.

340

& if that’s not exciting enough for you, I’ve got a great dog who’s been training to be my future lab’s lab (she has an impressive CV to show for it).

140

We'll be pursuing all things computational cognitive neuroscience, with focus on the neurocomputational principles of human learning & applications in brain-computer interfaces, neurofeedback, & mental health.

140

AND: in Fall 2027, I'll be starting my own lab at Vanderbilt University's College of Connected Computing as an Assistant Professor in the Computational Psychology area — a perfect spot for the strange and wonderful niche I've carved out spanning cog neuro, ML, applied mathematics, and psych.

150

I defended my PhD & I'm heading to Korea to join the Center for Neuroscience Imaging Research at Sungkyunkwan University as an IBS Young Scientist Fellow! Excited to explore new frontiers in neurofb alongside CNIR's brilliant scientists @hakwan.bsky.social @neuroaure.bsky.social @qilin7.bsky.social

2171